Research On Surface Nuclear Magnetic Resonance Configuration For Fine Detection Of Shallow Groundwater

Surface nuclear magnetic resonance（SNMR）is a geophysical method for groundwater detection based on nuclear magnetic resonance effect of hydrogen nuclei in groundwater.The earliest application of SNMR method was developed and implemented by scholars from the former Soviet Union in the 1980s in order to solve geological issues related to water resources exploration and large-scale hydrogeological survey.Compared with other geophysical methods for indirect groundwater exploration based on physical property such as resistivity,SNMR,due to its advantages,is more concerned by geophysicists in many groundwater exploration problems.In recent years,with the rapid development of technology and instruments,some geophysicists and scholars try to apply SNMR to some geological issues related to the detection of shallow groundwater.In the researches mentioned above,the target aquifer is shallower in most cases（depths are generally less than 30m）,but the division of aquifers and the estimation of water volume are required to be more precise.Although the detection of the traditional pulse moments observation mode（generally using 16 exponential pulse moment sequence）can effectively identify the underground aquifer region,the results are often limited to solve the relevant geological issues due to the lack of accuracy.In recent years,improving SNMR method from coil configuration and pulse moment configuration has become a hot topic.In the problems concerned shallower aquifer detection,the deployment of smaller transmitting coils at a lower cost may be considered.With the development of multichannel instrumentation,more complex coil loop configurations can be used.At the same time,the rapid development of machine learning and artificial intelligence provides a new idea for intelligent configuration of SNMR method pulse moment.The improvement of these schemes provides more possibilities for fine detection of shallow groundwater fundamentally.In view of the characteristics of shallow groundwater problems,based on exploration requirements and existing equipment conditions,this thesis extended the existing observation mode of SNMR method,proposed a set of method that can be used to screen and optimize the observation mode for specific exploration needs,and summarized the optimization observation mode for different shallow groundwater problems.By introducing the Bayesian optimization theory in machine learning and analyzing the characteristics of SNMR pulse moment observation method,a dynamic configuration of pulse moment based on Bayesian optimization was proposed.The combination of optimized observation mode and dynamic configuration of pulse moment based on Bayesian optimization constitutes a fine detection method to solve the shallow groundwater detection problem,which significantly improves the detection effect of the method.The main research work and achievements are as follows:1.Based on the above research background,this thesis carries out on the SNMR for fine detection scheme and method of shallow groundwater.It includes extending better multiple-loops observation scheme and developing new method of pulse moment configuration.The expansion of observation mode focuses on the needs of groundwater problems and the improvement of existing instruments and equipment,On the basis of considering the feasibility of the scheme,four multiple-loops observation configurations are proposed,namely the“multiple central-loops（coincident loop）","multiple central-loops（separated loop）”,“multiple gradient-loops（coincident loop）”,“multiple gradient-loops（separated loop）”.And four combined observation modes formed when the multiple-loops observation configuration is combined with the pulse moment observation configuration.2.A large number of studies have shown that the reasonable configuration of pulse moment is an important way to improve the resolution of the SNMR method.In this thesis,a novel sequential decision allocation method for pulse moments is proposed and implemented by introducing the Bayesian optimization method in machine learning.In this method,the actual observed data（E₀-q）curve is assumed to be an unknown objective function（“black box”function）.After each impulse moment observation,the objective function is regressed by Gaussian process to obtain the posterior distribution of the objective function under the existing observation data.The setting of the next pulse moment is suggested by the results obtained through“acquisition function”recommendation.In this thesis,a new covariance function kernel for SNMR method is developed and an acquistion function more suitable for SNMR method is designed.Compared with previous research results,the Bayesian optimal configuration method of pulse moment takes into account the increase of prior information in the process of impulse moment observation,The pulse moment sequence can be dynamically updated according to the existing observation data during the observation process,which can significantly improve the detection effect of the pulse moment observation mode under the condition of almost no increase in detection cost.3.Three typical shallow aquifer model are set up in this thesis,aiming at the problems including humidity detection in the cultural heritage protection where the target aquifer is mainly distributed in 0～10m,the sliding zone detection in landslide disaster where target aquifer is mainly distributed in 20～30m and the hydrogeological survey of water-bearing target is widely distributed in the shallow groundwater of0～30m.The inversion results of simulation data show that compared with traditional observation methods,the detection method proposed in this thesis can obviously improve the stratification effect of shallow groundwater inversion model and the accuracy of estimation of aquifer content.Finally,in order to verify the application effect of the detection method proposed in this thesis,it is applied to two site tests in Maqu County,Gannan Tibetan Autonomous Prefecture and Tianmen City,Hubei Province.The test results show that compared with the traditional observation methods,the fine detection method of shallow groundwater proposed in this thesis can effectively improve the detection effect of shallow groundwater.The above research results provide a theoretical basis and method technology for the research of SNMR detection scheme and detection method,and the research results have certain theoretical and application value.It provides a more precise detection scheme and method for SNMR detection of shallow groundwater.